| This thesis builds upon the work of Dr. Al Mundy and furthers the development of physicsbased reduced-order models of electric double-layer capacitors (EDLCs), by extending the operating range over which the models are accurate and by adding a thermal model. Electrical power management systems are often required either to sink or source large currents, which over the years has led to many solutions. Often the solution implemented for sinking is application of shunting resistors, which absorb the energy and then dissipate it as heat. Such a method wastes the entirety of the shunted energy. EDLCs offer an effective and efficient way to store charge and then source such large amounts of current in a manner that improves a system's efficiency. They have applications in regenerative braking for electric vehicles, efficient power regulation in modular power grid integration, and high-current sources for maglev trains or high-energy weapons such as lasers or rail-guns. Such systems require accurate and stable modeling to protect both the power distribution system, its components and the device. There are several methods for modeling EDLCs. These include equivalent circuits, which are easy to implement but do not offer knowledge of the internal state of the EDLC, neural networks which are a biologically inspired modeling scheme that allows for real-time adaptive machine learning, and finally, physics based reduced-order modeling, which provides accurate and rapid modeling of the EDLC's voltage and internal variables. Knowledge of these internal variable can be used to predict the precursors of premature degradation, and hence can be used by controls in an EDLCmanagement system to protect the device. This is the type of model that this thesis is focused on. In particular, this work extends the operating range over which the models are accurate and adds a thermal model. |
